Metaverse is a platform that offers unique user experiences. Users join the virtual worlds by using a virtual representation called an avatar. There is increasing use of NonFungible Token (NFT) avatars in metaverse applications that enable users to interact with each other in virtual worlds. These avatars contain not only personal information but also the behavioural footprints of users. Therefore, privacy preservation is critical for establishing a safe environment and enhancing the level of confidence in the metaverse. This paper proposes a novel privacy awareness framework that leverages machine learning (ML) algorithms to identify malicious patterns and alert users. This enables user awareness and the ability to adopt a required level of risk mitigation strategies. To test the framework, we have designed and implemented a test-bed where we can deploy various attack scenarios and collect data. We also present a case study on NFT cloning with the associated data collection.